Columbia economist warns AI infrastructure is crucial to US economic stability #
Without the hundreds of billions of dollars pouring into AI data centres, equivalent to more than £200 billion at current exchange rates, America would already be in a downturn. That is the stark warning from a Columbia Business School economist whose research has rippled through financial markets this month.
Stijn Van Nieuwerburgh, a real estate and finance professor at Columbia, argues that AI infrastructure spending has become the single biggest driver of US economic growth. 'We'd basically be in a recession right now without data centers,' he says. 'That's how big AI is.' His findings, first published by Columbia Business School, come as bond markets show growing unease about how all this construction is actually being paid for.
A Boom Bigger Than the Railroads #
Van Nieuwerburgh's research, titled 'Financing the AI Buildout', puts a number on the scale of the spending. He estimates AI infrastructure investment could average roughly 2.8% of GDP during the current buildout, larger than the railroad boom at its peak.
For context, railroad investment between 1865 and 1890 peaked at around 2.4% of GDP, while the postwar highway system and the 1990s telecoms fibre rollout never came close. The reason AI is so much more expensive, he explains, comes down to raw power. A traditional data-centre server rack might consume 5 to 10 kilowatts of power, whereas a rack filled with cutting-edge AI chips can require more than 100 kilowatts, forcing operators to install liquid cooling systems and dedicated electrical substations that a normal warehouse would never need.
Wall Street Starts Asking Who Is Really Holding the AI Debt #
That price tag, now running to upward of $8 billion for a single state-of-the-art AI campus, is increasingly being funded through debt rather than cash reserves. Van Nieuwerburgh points to one Meta-linked data centre project financed with roughly $27 billion in debt through a maze of special-purpose vehicles, a structure that keeps the borrowing off Meta's own balance sheet.
Bond investors appear to be catching on. Credit spreads for Google, Amazon and Meta have widened as fixed-income investors demand more reward to lend to the companies building out AI infrastructure. Oracle's five-year credit default swap, seen as a proxy for AI debt fears, is currently trading at a multi-year high.
Debt Hidden in the System #
Van Nieuwerburgh's central concern is not that the spending is wasteful, but that the risk is becoming harder to see. Because hyperscalers increasingly lease facilities rather than own them outright, pension funds, private credit firms and structured-finance vehicles now hold growing claims on AI infrastructure, often several steps removed from the tech companies actually using it.
'It's very hard to get the timing right with these big buildouts, and often what ends up happening is we get overexcited and accrue too much debt and then a bunch of these investments go bust,' he says. 'That doesn't mean these are not important, productive investments. What I'm arguing is that we need to be careful and think through the risks.'
Van Nieuwerburgh's research lands the same week that credit markets flashed a starker warning of their own. Credit default swap traders have priced roughly a 50% five-year default probability for CoreWeave, one of the largest 'neocloud' AI infrastructure providers, after its CDS spread topped around 855 basis points in late July, according to market data.
If enthusiasm for generative AI cools before these facilities generate the returns investors expect, the fallout would not stay confined to Silicon Valley. Pension funds, mortgage-bond holders and everyday savers with exposure to corporate credit markets could all feel the effects, making the health of the AI buildout a concern that stretches well beyond the tech sector itself. © Copyright IBTimes 2025. All rights reserved.